Classifying skies from images: A multidimensional approach to detecting high dynamic range imaging attributes

Abstract

Although the development of high dynamic range images allowed broadening sky vault analysis for lighting research purposes, the sky classification is still performed using subjective methods. There is no established metric for comparison, only isolated approaches that do not address the characteristics of the images. This paper presents a proposed multidimensional method for analysing light from sky vault high dynamic range images. A Matlab routine was applied. It uses a classification system to check the relevance of each image attribute, choosing the most suitable sky model from the ISO 15469:2004 (CIE S 011/E:2003) standard. The results, that can be plotted or exported, indicate that the routine is able to choose the most relevant model for the photographed sky, thus allowing the creation of a sky classification database.